def chunk_file_iterator(chunk_samples, fs, seiz_file):
n = 0
try:
while True:
data, _, labels = edfread(
seiz_file,
rec_times=(n * chunk_samples / float(fs),
(n + 1) * chunk_samples / float(fs)))
#data_is_full = np.size(data,0)>=chunk_samples
start_time = n * chunk_samples / float(fs)
yield data, start_time, n, labels
n += 1
except ValueError:
pass
def chunk_file_iterator(chunk_samples,fs,seiz_file, good_channels = None):
"""
iterates over chunks of data. Useful for long (longer than 5 minutes) edf files, saves memory.
"""
n = 0
try:
while True:
data,_,labels = edfread(seiz_file,
rec_times=(n*chunk_samples/float(fs),(n+1)*chunk_samples/float(fs)),
good_channels=good_channels)
#data_is_full = np.size(data,0)>=chunk_samples
start_time = n*chunk_samples/float(fs)
yield data, start_time,n, labels
n+=1
except ValueError:
pass
def parallel_coherence(file_path, win_len, win_overlap, fs, nprocs, save_path):
# these are hard-coded parameters to
# chunk up data into 10 min chunks
min_per_chunk = 10
sec_per_min = 60
i = 0
while True:
# get chunk start and end times
start = i * sec_per_min * min_per_chunk
end = (i+1) * sec_per_min * min_per_chunk + float(win_overlap) / fs
# get the chunk
try:
# extract the chunk
print 'Extracting the ' + str(i) + ' chunk'
X_chunk, _, labels = edfread(file_path, rec_times = [start, end])
print 'Printing the labels:'
for i in range(len(labels)):
print '\tChannel ' + str(i) + ' is ' + labels[i]
# compute coherence for this chunk and save json file
this_save_path = save_path + ".json"
mp_coherence(X_chunk, win_len, win_overlap, fs, nprocs, this_save_path)
# if less than an entire chunk was read, then this is the last one!
if X_chunk.shape[0] < sec_per_min * min_per_chunk:
break
except ValueError:
break # the start was past the end!
return
def analyze_patient(data_path, save_path, patient_id, res_f, window_length=1.0, window_overlap=0.5, num_windows=3000, f_s=1e3, include_awake=True, include_asleep=False):
# reformat window length and overlap as indices
window_length = int(window_length * f_s)
window_overlap = int(window_overlap * f_s)
# create save path
if not os.path.isdir(save_path):
os.makedirs(save_path)
# specify data paths
print 'Specifying file paths'
if not os.path.isdir(data_path):
sys.exit('Error: Specified data path does not exist')
p_file = os.path.join(data_path, 'patient_pickle.txt')
with open(p_file,'r') as pickle_file:
patient_info = pickle.load(pickle_file)
# add data file names
data_filenames = patient_info['seizure_data_filenames']
seizure_times = patient_info['seizure_times']
con_type = ['ictal'] * len(data_filenames)
if include_awake:
data_filenames += patient_info['awake_inter_filenames']
seizure_times += [None] * len(patient_info['awake_inter_filenames'])
con_type += ['awake'] * len(patient_info['awake_inter_filenames'])
if include_asleep:
data_filenames += patient_info['asleep_inter_filenames']
seizure_times += [None] * len(patient_info['asleep_inter_filenames'])
con_type += ['sleep'] * len(patient_info['asleep_inter_filenames'])
data_filenames = [os.path.join(data_path,filename) for filename in data_filenames]
num_files = len(data_filenames)
# get data in numpy array
print 'Reading data from edf files to numpy array'
all_data = []
num_channels = []
i = 1
for seizure_file in data_filenames:
print '\tReading ' + str(i) + ' of ' + str(num_files)
i += 1
X,_,_ = edfread(seizure_file)
num_channels.append(X.shape[1])
all_data.append(X)
if len(set(num_channels)) == 1:
num_channels = num_channels[0]
gt1 = num_channels > 1
print 'There ' + 'is '*(not gt1) + 'are '*gt1 + str(num_channels) + ' channel' + 's'*gt1
else:
print 'Channels: ' + str(num_channels)
sys.exit('Error: There are different numbers of channels being used for different seizure files...')
# get the number of parameters (3 energy statistics per channel)
p_feat = 3 * num_channels
# pre-process data -- filter parameters
print 'Applying a band-pass filter to the data'
band = np.array([0.1,100.])
band_norm = band / (f_s / 2.) # normalize the band
filt_order = 3
# band pass filter the data
b, a = signal.butter(filt_order, band_norm, 'bandpass') # design filter
for j in range(num_files):
all_data[j] = signal.filtfilt(b,a,all_data[j],axis=0) # filter the data
# run leave-one-out cross validation testing
sensitivity, latency, FP, time = loocv_testing(all_data, con_type, window_length, window_overlap, num_windows, f_s, seizure_times, p_feat, save_path)
# get mean statistics
m_sense = np.nanmean(sensitivity)
m_latency = np.nanmean(latency)
m_fpr = np.nansum(FP) / np.nansum(time)
# print to results file
print >> res_f, '\nPatient ' + patient_id + '\n========================='
# print the results -- aggregates and total
print >> res_f, 'Mean Sensitivity: \t%.2f' %(m_sense)
print >> res_f, 'Mean Latency: \t%.4f' %(m_latency)
print >> res_f, 'False Positive Rate: \t%.5f (fp/Hr) \n' % m_fpr
print >> res_f, 'Sensitivity: ' + str(sensitivity)
print >> res_f, 'Latency: ' + str(latency)
print >> res_f, 'False Positive Rate: ' + str(FP / time)
return sensitivity, latency, m_fpr
while True:
# get chunk start and end times
start = desired_chunk_index * sec_per_min * min_per_chunk
end = (desired_chunk_index + 1) * sec_per_min * min_per_chunk
print '\tStart time in seconds:', start
print '\tEnd time in seconds:', end
try:
# extract the chunk
print '\t\t\tChunk ' + str(ind) + ' reading...\n',
# Gotta get the data_filename right!
dimensions_to_keep = choose_best_channels(
patient_id, seizure=0, filename=data_filename_path)
X_chunk, _, labels = edfread(data_filename_path,
rec_times=[start, end],
good_channels=dimensions_to_keep)
# Added: if the readed chunk is too short, break.
if X_chunk.shape[0] < 300 * f_s:
print "Chunk is too short! Pick different chunk to analyze!"
break
# update file information
all_files.append(X_chunk)
tmp_data_filenames = data_filename_path
tmp_file_type = file_type
tmp_seizure_times = seizure_times
print '\t\t\tInterictal chunk %d reading complete!' % (ind)
break
except ValueError:
def analyze_patient_raw(data_path,
f_s=1e3,
include_awake=True,
include_asleep=False,
long_interictal=False):
# minutes per chunk (only for long interictal files)
min_per_chunk = 15
sec_per_min = 60
# specify data paths
if not os.path.isdir(data_path):
sys.exit('Error: Specified data path does not exist')
p_file = os.path.join(data_path, 'patient_pickle.txt')
with open(p_file, 'r') as pickle_file:
print("Open Pickle: {}".format(p_file) + "...\n")
patient_info = pickle.load(pickle_file)
# # add data file names
data_filenames = patient_info['seizure_data_filenames']
seizure_times = patient_info['seizure_times']
file_type = ['ictal'] * len(data_filenames)
seizure_print = [True] * len(data_filenames) # mark whether is seizure
if include_awake:
data_filenames += patient_info['awake_inter_filenames']
seizure_times += [None] * len(patient_info['awake_inter_filenames'])
file_type += ['awake'] * len(patient_info['awake_inter_filenames'])
seizure_print += [False] * len(patient_info['awake_inter_filenames'])
if include_asleep:
data_filenames += patient_info['asleep_inter_filenames']
seizure_times += [None] * len(patient_info['asleep_inter_filenames'])
file_type += ['sleep'] * len(patient_info['asleep_inter_filenames'])
seizure_print += [False] * len(patient_info['asleep_inter_filenames'])
data_filenames = [
os.path.join(data_path, filename) for filename in data_filenames
]
good_channels = patient_info['good_channels']
# band pass filter parameters
# band = np.array([0.1, 100.])
# band_norm = band / (f_s / 2.) # normalize the band
# filt_order = 3
# b, a = signal.butter(filt_order, band_norm, 'bandpass') # design filter
# get data in numpy array
num_channels = []
all_files = []
all_files_unfiltered = []
tmp_data_filenames = []
tmp_file_type = []
tmp_seizure_times = []
tmp_seizure_print = []
print 'Getting Data...'
for i, seizure_file in enumerate(data_filenames):
# this is for when we have inter-ictal files that are an hour long that has split it up into parts
if long_interictal and not (file_type[i] is 'ictal'):
print '\tThis code has not been written'
else:
print '\tSeizure file %d reading...' % (i + 1),
# read data in
X, _, labels = edfread(seizure_file)
all_files_unfiltered.append(X)
n, p = X.shape
num_channels.append(p)
# good_channels_ind = []
# labels = list(labels)
# for channel in good_channels:
# good_channels_ind.append(labels.index(channel))
# # filter data
# print 'filtering...',
# X = signal.filtfilt(b, a, X, axis=0) # filter the data
all_files.append(X) # add raw data to files
# update temporary stuff
data_filenames = update_list(data_filenames, tmp_data_filenames)
file_type = update_list(file_type, tmp_file_type)
seizure_times = update_list(seizure_times, tmp_seizure_times)
seizure_print = update_list(seizure_print, tmp_seizure_print)
# double check that the number of channels matches across data
if len(set(num_channels)) == 1:
num_channels = num_channels[0]
gt1 = num_channels > 1
print 'There ' + 'is ' * (not gt1) + 'are ' * gt1 + str(
num_channels) + ' channel' + 's' * gt1 + "\n"
else:
print 'Channels: ' + str(num_channels)
print 'There are inconsistent number of channels in the raw edf data'
sys.exit(
'Error: There are different numbers of channels being used for different seizure files...'
)
# double check that no NaN values appear in the features
for X, i in enumerate(all_files):
if np.any(np.isnan(X)):
print 'There are NaN in raw data of file', i
sys.exit('Error: Uh-oh, NaN encountered while extracting features')
return all_files, data_filenames, file_type, seizure_times, seizure_print
def analyze_patient(patient_id, data_path, save_path, log_file, parameters, folds, win_len=1.0, win_overlap=0.5, num_windows=1000, f_s=1e3,include_awake=True, include_asleep=False, long_interictal=False):
# minutes per chunk (only for long interictal files)
min_per_chunk = 15
sec_per_min = 60
# reformat window length and overlap as indices
win_len = int(win_len * f_s)
win_overlap = int(win_overlap * f_s)
# create save path
if not os.path.isdir(save_path):
os.makedirs(save_path)
# specify data paths
if not os.path.isdir(data_path):
sys.exit('Error: Specified data path does not exist')
p_file = os.path.join(data_path, 'patient_pickle.txt')
with open(p_file,'r') as pickle_file:
print("Open Pickle: {}".format(p_file)+"...\n")
patient_info = pickle.load(pickle_file)
# add data file names
data_filenames = patient_info['seizure_data_filenames']
seizure_times = patient_info['seizure_times']
file_type = ['ictal'] * len(data_filenames)
seizure_print = [True] * len(data_filenames) # mark whether is seizure
if include_awake:
data_filenames += patient_info['awake_inter_filenames']
seizure_times += [None] * len(patient_info['awake_inter_filenames'])
file_type += ['awake'] * len(patient_info['awake_inter_filenames'])
seizure_print += [False] * len(patient_info['awake_inter_filenames'])
if include_asleep:
data_filenames += patient_info['asleep_inter_filenames']
seizure_times += [None] * len(patient_info['asleep_inter_filenames'])
file_type += ['sleep'] * len(patient_info['asleep_inter_filenames'])
seizure_print += [False] * len(patient_info['asleep_inter_filenames'])
data_filenames = [os.path.join(data_path,filename) for filename in data_filenames]
good_channels = patient_info['good_channels']
# band pass filter parameters
band = np.array([0.1,100.])
band_norm = band / (f_s / 2.) # normalize the band
filt_order = 3
b, a = signal.butter(filt_order, band_norm, 'bandpass') # design filter
# get data in numpy array
num_channels = []
all_files = []
tmp_data_filenames = []
tmp_file_type = []
tmp_seizure_times = []
tmp_seizure_print = []
print 'Getting Data...'
for i, seizure_file in enumerate(data_filenames):
# check that we haven't already gotten energy statistics for this seizure file
file_begin = os.path.join(save_path, os.path.splitext(os.path.basename(data_filenames[i]))[0])
# this is for when we have inter-ictal files that are an hour long that has split it up into parts
if long_interictal and not (file_type[i] is 'ictal'):
# get all files in the save path
all_files_in_dir = [os.path.join(save_path,fn) for fn in next(os.walk(save_path))[2]]
# all_files_in_dir = os.listdir(save_path)
# if the energy stats exist for this file(.es represents energy statistics)
inter_ictal_files = [s for s in all_files_in_dir if s.startswith(file_begin) and s.endswith(".es")]
# if inter ictal (energy) files found, read them
if inter_ictal_files:
tmp_data_filenames.append( (i, inter_ictal_files))
tmp_file_type.append( (i, [file_type[i]] * len(inter_ictal_files)) )
tmp_seizure_times.append( (i, [seizure_times[i]] * len(inter_ictal_files)) )
tmp_seizure_print.append( (i, [seizure_print[i]] * len(inter_ictal_files)) )
# read each of the interictal files
print "\tSeizure file %d is long --reading energy statistics directly..." %(i+1),
for j, file_name in enumerate(inter_ictal_files):
print "%d" %(j+1),
X_feat, p = read_energy_file(file_name)
all_files.append(X_feat) # feature vector X from all the inter_ictal_files saved to all_files
num_channels.append(p)
print " "
else:
# read the file
print '\tSeizure file %d is long...' %(i+1)
j = 0
while True:
# get chunk start and end times
start = j * sec_per_min * min_per_chunk
end = (j+1) * sec_per_min * min_per_chunk
# get the chunk
try:
# extract the chunk
print '\t\tChunk ' + str(j+1) + ' reading...',
X_chunk, _, labels = edfread(seizure_file, rec_times = [start, end])
n,p = X_chunk.shape
num_channels.append(p)
good_channels_ind = []
labels = list(labels)
for channel in good_channels:
good_channels_ind.append(labels.index(channel))
# compute feature vector
print 'filtering...',
X_chunk = signal.filtfilt(b,a,X_chunk,axis=0) # filter the data
# get feature vectors from windows -- energy statistics
print 'extracting features...',
n_windows = n / (win_len - win_overlap) - 1 # evaluates to floor( n / (L - O ) - 1 since ints
X_feat = np.empty((n_windows,3,p))
m = 0
for k in range(win_len, X_chunk.shape[0], win_len - win_overlap):
window = X_chunk[(k-win_len):k,:] # select window
f = energy_features(window) # extract energy statistics
X_feat[m,:,:] = f
m += 1
all_files.append(X_feat) # add feature to files
# save energy statistics file
print 'saving...'
es_file = file_begin + "_%d.es"%(j)
write_energy_file(es_file, X_feat)
# print to csv if this is a desired seizure file
if seizure_print[i]:
filtered_file = file_begin + '_%d_filtered.csv'%(j)
energy_file = file_begin + '_%d_energystats.csv'%(j)
create_filtered_csv(filtered_file, X_chunk, good_channels_ind)
create_energy_csv(energy_file, X_feat[:,:,good_channels_ind[0]])
# update count
j += 1
# if less than an entire chunk was read, then this is the last one!
if X_chunk.shape[0] < sec_per_min * min_per_chunk:
break
except ValueError:
print "no wait, that doesn't exist!"
break # the start was past the end!
# store temporary stuff
tmp_data_filenames.append( (i, [os.path.join(save_path,file_begin + "_%d.es"%(k)) for k in range(j)]) )
tmp_file_type.append( (i, [file_type[i]] * j) )
tmp_seizure_times.append( (i, [seizure_times[i]] * j) )
tmp_seizure_print.append( (i, [seizure_print[i]] * j) )
else:
es_file = file_begin + ".es"
if os.path.isfile(es_file):
print "\tSeizure file %d --reading energy statistics directly" %(i+1)
X_feat, p = read_energy_file(es_file)
all_files.append(X_feat)
num_channels.append(p)
else:
print '\tSeizure file %d reading...' %(i+1),
# read data in
X,_,labels = edfread(seizure_file)
n,p = X.shape
num_channels.append(p)
good_channels_ind = []
labels = list(labels)
for channel in good_channels:
print "channel",channel
good_channels_ind.append(labels.index(channel))
# filter data
print 'filtering...',
X = signal.filtfilt(b,a,X,axis=0) # filter the data
# get feature vectors from windows -- energy statistics
print 'extracting features...'
n_windows = n / (win_len - win_overlap) - 1 # evaluates to floor( n / (L - O ) - 1 since ints
X_feat = np.empty((n_windows,3,p))
k = 0
for j in range(win_len, X.shape[0], win_len - win_overlap):
window = X[(j-win_len):j,:] # select window
f = energy_features(window) # extract energy statistics
X_feat[k,:,:] = f
k += 1
all_files.append(X_feat) # add feature to files
# save energy statistics file
write_energy_file(es_file, X_feat)
# print to csv if this is a desired seizure file
if seizure_print[i]:
filtered_file = file_begin + '_filtered.csv'
energy_file = file_begin + '_energystats.csv'
create_filtered_csv(filtered_file, X, good_channels_ind)
create_energy_csv(energy_file, X_feat[:,:,good_channels_ind[0]])
# update temporary stuff
data_filenames = update_list(data_filenames, tmp_data_filenames)
file_type = update_list(file_type, tmp_file_type)
seizure_times = update_list(seizure_times, tmp_seizure_times)
seizure_print = update_list(seizure_print, tmp_seizure_print)
# double check that the number of channels matches across data
if len(set(num_channels)) == 1:
num_channels = num_channels[0]
gt1 = num_channels > 1
print 'There ' + 'is '*(not gt1) + 'are '*gt1 + str(num_channels) + ' channel' + 's'*gt1+"\n"
else:
print 'Channels: ' + str(num_channels)
sys.exit('Error: There are different numbers of channels being used for different seizure files...')
# double check that no NaN values appear in the features
for X in all_files:
if np.any(np.isnan(X)):
sys.exit('Error: Uh-oh, NaN encountered while extracting features')
# leave one out cross validation, update log
fitnesses=loocv_testing(all_files, data_filenames, file_type, seizure_times, seizure_print, win_len, win_overlap, num_windows, f_s, save_path,parameters, folds)
#update_log(log_file, patient_id, sensitivity, latency, FP, time)
return fitnesses
# # # asleep_file= "/Users/TianyiZhang/Documents/EpilepsyVIP/data/TS041/DA00101Q_1-1_02oct2010_03_00_05_Sleep+.edf" # # X_asleep, _,labels_asleep = edfread(asleep_file) # # print "how many channels for asleep file?",len(labels_asleep) # # awake_file= "/Users/TianyiZhang/Documents/EpilepsyVIP/data/TS041/DA00101P_1-1_02oct2010_09_00_38_Awake+.edf" # # X_awake, _,labels_awake = edfread(awake_file) # # print "how many channels for asleep file?",len(labels_awake) file = "/Users/TianyiZhang/Documents/EpilepsyVIP/data/TS039/CA00100D_1-1+.edf" X, _, labels = edfread(file) expected = "RAH3" print labels.index(expected) """Test Filters""" # # Filter a noisy signal. # T = 0.05 # nsamples = T * fs # t = np.linspace(0, T, nsamples, endpoint=False) # a = 0.02 # f0 = 600.0 # x = 0.1 * np.sin(2 * np.pi * 1.2 * np.sqrt(t)) # x += 0.01 * np.cos(2 * np.pi * 312 * t + 0.1) # x += a * np.cos(2 * np.pi * f0 * t + .11) # x += 0.03 * np.cos(2 * np.pi * 2000 * t) # plt.figure(2)
def classif_data_collect(seiz_filenames,
seizure_times,
inter_filenames,
window_len,
preictal_time,
postictal_time,
n_windows,
sliding_window=False,
window_overlap=.8,
fs=1000.,
good_channels=None,
bad_channels=None,
rstat_bands=((1, 4), (5, 8), (9, 13), (14, 25),
(25, 90), (100, 200)),
rstat_win_len=20000,
notch_filt=True,
norm_whole_file=True,
norm_window=False):
'''
:param seiz_filenames: A list of the .edf filenames which contain epileptic data
:param seizure_times: A list corresponding to the ictal filenames of:
tuples containing seizure start and end times (in seconds),
if the recording contains of seizure
For instance, if my filename list looks like
['seizure_a.edf','seizure_b.edf']
Then my seizure_times list might be:
[(123,135),(60,300)]
:param inter_filenames: A list of .edf filenames of interictal data
:param good_channels: A list of the channel names that should be observed for the
:param window_len: The length of the windows that the program takes,
in number of samples
:param preictal_time: The amount of time, in seconds, before a seizure, for which
a window will still be considered as preictal for training purposes
:param postictal_time:The amount of time, in seconds, after a seizure
that the period is considered postictal
:param n_windows: The number of windows of each class we choose to extract
:param sliding_window: Do you want the data to be a sliding window?
:param window_overlap: The maximum allowable overlap (percentage) between two windows.
:param fs: the sampling frequency of the sample
:param bad_channels: if you want to read in all channels, except for a select few
:param rstat_bands: An iterable of pairs of two elements,
representing the start and stop of the bands of interest. Can be NoneType if no rstat is desired
:param rstat_win_len: (int) The length of the random windows
extracted in computing the r-statistic
:param notch_filt: (bool) A boolean value determining whether a notch filter is to be applied to the data
:param norm_whole_file: (bool) A boolean value that normalizes the whole file, not just the window.
:param norm_window: (bool) A boolean value that chooses to normalize each window, not the whole file.
:return:
'''
def select_windows_from_interval(int_start, int_end, max_iter=10):
'''
:param int_start:
:param int_end:
:param max_iter:
:return: window_ends, a numpy array with indices at which the sample ends
'''
window_ends = np.sort(
np.random.randint(int_start + window_len, int_end, n_windows))
if n_windows == 1:
return window_ends
for i in range(max_iter):
diffs = np.diff(window_ends)
if all(diffs >= (window_len * window_overlap)):
return window_ends
#TODO: refine window logic
window_ends = np.sort(
np.random.randint(int_start + window_len, int_end, n_windows))
else:
#manually select window ends
window_ends = int_end - window_overlap * window_len * np.arange(
n_windows)
return window_ends
def generate_windows_from_seizure(sliding_window=False):
'''
Given seizure data, select random windows
from the preictal,ictal,and postictal phases
:return:
'''
preictal_samples, postictal_samples = int(preictal_time * fs), int(
postictal_time * fs)
seiz_start, seiz_end = seize_times
ictal_samples = int((seiz_end - seiz_start) * fs)
max_iter = 4
if not sliding_window:
if not window_overlap:
if (n_windows*window_len) > \
(min(preictal_samples, postictal_samples,ictal_samples)):
raise ValueError(
'Nonoverlapping windows not possible to produce')
else:
if (n_windows*window_len - (n_windows-1)*window_len*window_overlap)>\
min(preictal_samples,postictal_samples,ictal_samples):
raise ValueError(
'Not possible to produce overlapping windows with certain max_proportion'
)
seiz_start = int(seiz_start * fs)
seiz_end = int(seiz_end * fs)
interval_starts = (seiz_start - preictal_samples, seiz_start, seiz_end)
interval_ends = (seiz_start, seiz_end, np.size(seiz_data, axis=0))
labels = ['Preictal', 'Ictal', 'Postictal']
if sliding_window:
end_times = range(window_len - 1, np.size(seiz_data, 0),
int(window_len * (1 - window_overlap)))
for end in end_times:
start = end - window_len + 1
lab_index = interval_inclusion_index(start, end,
interval_starts,
interval_ends)
if lab_index is None:
label = 'None of the Above'
else:
label = labels[lab_index]
window = preprocess(seiz_data[start:end + 1, :],
normaliz=norm_window,
notch_filt=False)
data_container.append({
'window': window,
'label': label,
'fold_lab': 'S{}'.format(seiz_count),
'time': end / fs
})
else:
for interval_start, interval_end, label in zip(
interval_starts, interval_ends, labels):
window_ends = select_windows_from_interval(
interval_start, interval_end, max_iter)
for end in window_ends:
window = preprocess(seiz_data[end - window_len + 1:end +
1, :],
normaliz=norm_window,
notch_filt=False)
data_container.append({
'window': window,
'label': label,
'fold_lab': 'S{}'.format(seiz_count),
'time': end / fs
})
return
def generate_windows_from_nonseizure(sliding_window=False):
label = 'Interictal'
if sliding_window:
end_times = range(window_len, np.size(seiz_data, 0),
int(window_len * (1 - window_overlap)))
for end in end_times:
window = preprocess(seiz_data[end - window_len + 1:end + 1, :],
normaliz=norm_window,
notch_filt=False)
data_container.append({
'window': window,
'label': label,
'fold_lab': 'NS{}'.format(non_seiz_count),
'time': end / fs
})
else:
max_iter = 4 # maximum times to pick a random list for nonoverlapping windows
end_times = select_windows_from_interval(
0, np.size(seiz_data, axis=0), max_iter)
for end in end_times:
window = preprocess(seiz_data[end - window_len + 1:end + 1, :],
normaliz=norm_window,
notch_filt=False)
data_container.append({
'window': window,
'label': label,
'fold_lab': 'NS{}'.format(non_seiz_count),
'time': end / fs
})
return
def preprocess(seiz_data,
normaliz=False,
notch_filt=True,
rstat_band=False,
causal=True):
if normaliz:
seiz_data = normalize(seiz_data)
if notch_filt:
seiz_data = notch(seiz_data, 56., 64., fs, mode=not causal)
if rstat_band:
seiz_data = rstat_processor.optimal_bandpass(seiz_data,
mode=not causal)
return seiz_data
###start main code of the function
rstat_filt = False
if rstat_bands is not None:
rstat_processor = RstatPreprocessor(inter_filenames[0],
seiz_filenames[0],
seizure_times=seizure_times[0],
fs=1000.)
rstat_processor.prepare_rstat(rstat_bands,
good_channels=good_channels,
bad_channels=bad_channels,
window_len=20000,
mode=0)
rstat_filt = True
data_container = []
if not window_overlap:
window_overlap = 0
for seiz_count, (seizure_file, seize_times) in enumerate(
zip(seiz_filenames, seizure_times)):
seiz_data, _, _ = edfread(seizure_file,
good_channels=good_channels,
bad_channels=bad_channels)
seiz_data = preprocess(seiz_data,
normaliz=norm_whole_file,
notch_filt=notch_filt,
rstat_band=rstat_filt)
generate_windows_from_seizure(sliding_window)
for non_seiz_count, nonseizure_file in enumerate(inter_filenames):
seiz_data, _, _ = edfread(nonseizure_file,
good_channels=good_channels,
bad_channels=bad_channels)
seiz_data = preprocess(seiz_data,
normaliz=norm_whole_file,
notch_filt=notch_filt,
rstat_band=rstat_filt)
generate_windows_from_nonseizure(sliding_window)
return {'data': data_container, 'seize_times': seizure_times}
os.getcwd()))) # direct to EpilepsyVIP
data_path = os.path.join(to_data, 'data') # direct to data folder
for i, patient_id in enumerate(patients):
# update paths specific to each patient
p_data_path = os.path.join(data_path, patient_id)
print "---------------------------Analyzing patient ", patient_id, "----------------------------\n"
# if data path does not work out
if not os.path.isdir(data_path):
sys.exit('Error: Specified data path does not exist')
# get pickle file
p_file = os.path.join(p_data_path, 'patient_pickle.txt')
# open pickle file and load
with open(p_file, 'r') as pickle_file:
print("Open Pickle: {}".format(p_file) + "...\n")
patient_info = pickle.load(pickle_file)
data_filenames = patient_info['seizure_data_filenames']
seizure_times = patient_info['seizure_times']
file_type = ['ictal'] * len(data_filenames)
seizure_print = [True] * len(data_filenames) # mark whether is seizure
print 'Getting Data...'
# read seizure file
for i, seizure_file in enumerate(data_filenames):
path_to_seizure = os.path.join(p_data_path, seizure_file)
print path_to_seizure
x, _, labels = edfread(path_to_seizure)
# output x, raw iEEG signal
print 'There are', x.shape[1], 'channels'
selected_features = feature_selection(x)
labels = clustering(selected_features)